In recent years, diets of widely varying composition have been used in the treatment of obesity. While none of these diets has produced consistent, long-term weight loss, some individuals on virtually all types of diets do remarkably well. One explanation for this variation among individuals is differences in motivation and behavior. However, underlying biological differences may also play an important role. Previous work has identified insulin secretion (as quantified by serum insulin concentration at 30 minutes following a standard 75-gram oral glucose load) as an important biological determinant. This randomized-controlled, multi-center trial aims to test the hypothesis that insulin secretion will predict which type of diet will work best for each individual. A total of 160 obese young adults (18 to 40 years) will be assigned to one of two diets: low-fat (60% carbohydrate, 20% fat, 20% protein) or low-glycemic load (45% carbohydrate, 35% fat, 20% protein). Participants will be enrolled at two sites (Children's Hospital Boston or University of North Carolina-Chapel Hill) and assigned to diet groups according to baseline insulin secretion status. The protocol will include an intensive 6-month intervention period and a 12-month follow-up period. Registered dietitians will provide nutrition education and behavioral counseling during group workshops and scheduled telephone calls. In an effort to avoid confounding, close attention will be paid to control for treatment intensity, physical activity prescriptions, and behavioral methods between groups. The primary endpoint will be percent body fat by DXA scan. Secondary endpoints will include cardiovascular and diabetes risk factors. Repeated, 24-hour dietary and physical activity recall interviews will provide process data. Metabolomic profiling and molecular analysis of identified metabolites will be conducted to characterize phenotype and explore potential physiological mechanisms.